Master Thesis in Humanitarian Action in Conflict Department of Theology Uppsala University Spring 2017, 15 Credits

“If you don’t take off your clothes, we’re going to kill you”

Sexual electoral violence as a silent weapon in Burundi and implications for humanitarian action

Miriam van Baalen Supervisor: Florian Krampe, Department of Peace and Conflict Research

This thesis is submitted for obtaining the Master’s Degree in International Humanitarian Action and Conflict. By submitting the thesis, the author certifies that the text is from his/her hand, does not include the work of someone else unless clearly indicated, and that the thesis has been produced in accordance with proper academic practices.

To the women of Burundi, who have been victims of sexual violence in war and peace. May the future bring justice.

2 ABSTRACT

Previous research on electoral violence has explored the variation in manifestation, perpetrators, victims and causes or motivations. While explanations have prescribed to structural and electoral factors or the nature of politics, nature of elections and electoral institutions, the causes and motivations behind sexual electoral violence has been under- researched. This in-depth single case study of sexual violence against women associated with the opposition in Burundi explores the possibilities of building bridges between theory on electoral violence and sexual and gender-based violence. In order to shed light over the disciplinary grey-zone, the study makes a theoretical contribution by suggesting a novel definition necessary to capture the phenomenon. The fusion of previous research suggests that similarly to wartime rape, sexual electoral violence can be used as an effective tool of political coercion. The argument is explored empirically with the case of Burundi, concluding that strategic rape is not exclusive to wartime contexts and under sociocultural conditions of high stigmatisation of sexual violence, it can be used as silent weapon of repression against the opposition. Reducing sexual electoral violence to general sexual violence is problematic for both policy-making and humanitarian action, as they require diverse action. The political dimension of sexual violence unravels the divide within the heart of humanitarian action between the classic Dunantist philosophy and its new, more progressive counterpart.

Keywords: Conflict, civil war, sexual and gender-based violence, electoral violence, sexual electoral violence, impunity, Burundi, Africa.

3 ABBREVIATIONS CNDD-FDD The National Council for the Defence of Democracy EV Electoral violence FNL National Forces of Liberation FRODEBU Front for Democracy in Burundi HRW Human Rights Watch SEV Sexual electoral violence SGBV Sexual and gender-based violence UPRONA Union for National Progress VAWIE Violence against women in elections VAWIP Violence against women in politics

4 TABLE OF CONTENTS 1. INTRODUCTION ...... 6 2. THEORY ...... 9 2.1 PREVIOUS RESEARCH ON ELECTORAL VIOLENCE ...... 9 2.2 FOCUSING ON THE DYNAMICS OF SEXUAL VIOLENCE ...... 12 2.3 TOWARDS CONCEPTUAL CLARITY ...... 15 3. RESEARCH DESIGN ...... 17 3.1 RESEARCH DESIGN ...... 17 3.2 CASE SELECTION ...... 17 3.3 STRUCTURE OF EMPIRICAL DESCRIPTION AND ANALYSIS ...... 18 3.4 MATERIAL AND DELIMITATIONS ...... 19 4. OBSERVATIONS OF VIOLENCE IN BURUNDI ...... 20 4.1 SGBV AND THE LEGACY OF CIVIL WAR ...... 20 4.2 ELECTORAL VIOLENCE IN BURUNDI AND THE CRISIS OF 2015 ...... 22 4.3 SEXUAL ELECTORAL VIOLENCE SURROUNDING THE ELECTORAL CRISIS OF 2015 ...... 24 5. CONTRASTING SEXUAL ELECTORAL VIOLENCE ...... 28 5.1 THE DYNAMICS OF VIOLENCE ...... 28 5.2 THE DISSIMILARITY OF SEXUAL ELECTORAL VIOLENCE ...... 28 5.3 ALTERNATIVE EXPLANATIONS AND CRITICAL REFLECTION ...... 30 5.4 IMPLICATIONS FOR POLICY AND HUMANITARIAN ACTION ...... 32 6. CONCLUSIONS ...... 34 6.1 MAIN CONCLUSIONS ...... 34 6.2 FUTURE RESEARCH ...... 34 BIBLIOGRAPHY ...... 36 APPENDIX A...... 42 CAUSES OF ELECTORAL VIOLENCE ...... 42 APPENDIX B...... 43 KEY EVENTS SURROUNDING THE 2015 ELECTORAL CRISIS ...... 43

5 1. INTRODUCTION

“If you don’t take off your clothes, we are going to kill you. (…) We’ll finish you all. Do you think when we’re done with your husbands and your brothers, you’ll remain? We have to eradicate you.”

Victim of targeted rape in Bujumbura, December 2015. (Quoted in Steers, 2016)

In July 2016, Human Rights Watch (HRW) published a shocking report documenting 323 cases of rape or sexual assault in Burundi between May and September 2015, appearing to specifically target family members of perceived government opponents. In many of the cases, the perpetrators where identified as loyal to the ruling-party, the Council for the Defense of Democracy – Forces of the Defense of Democracy (CNDD-FDD). This pattern of abuse is related to the political crisis that unfolded following the controversial re-election of President Pierre Nkurunziza for a third term in office (HRW, 2016a). While rape was used as a strategy during the civil war between 1993 and 2008 (Pézard and De Tessières, 2009, pp.76–81) the targeting of family members of the opposition in times of relative peace and in relation to elections is puzzling. Burundi is not alone with the problems of sexual and gender-based violence (SGBV) during war and peace. Documented cases of wartime rape in the Democratic Republic of Congo, Uganda, Burma and Bosnia have shown that rape can be a tool of political violence repressing political action, but is also used to repress, dishonour or humiliate political opponents. Rape as a form of electoral violence (EV) in non-conflict states, like Guinea and Côte d’Ivoire in 2010, shares the same effects as a high impact tool of political coercion (Bardall, 2015, p.17). However, the latter cases involved the targeting of women in their own political role, rather than the association to a political stakeholder as seen in Burundi. Thus, this raises the question: What are the causes of sexual electoral violence (SEV) against women associated to the perceived opposition? Does this form of violence differ from general EV and SGBV?

6 This study seeks to elucidate possible explanations of the observed sexual violence in Burundi, introducing the novel conceptual definition of SEV. An electoral process is supposed to be “an alternative to violence as it is a means of achieving government” (Fisher, 2002, p.2). Previous research on EV has made efforts to categorise the explanations after its causes and motivations. Höglund (2009) has divided the explanations into the nature of politics, nature of elections and electoral institutions. Some underlying conditions enable the use of EV, while certain factors trigger it. Similarly, Adolfo et al. (2012) have divided the causes into two broader categories, where structural factors or electoral factors are considered explanatory. However, EV involves a multitude of actors, motives and activities, which “renders simple explanations for its occurrence insufficient” (Höglund, 2009, p.424) and strategies of the actors involved “are also important in explaining the bearing of violence on electoral processes” (Höglund, 2009, p.424). While some efforts have been made to shed light gendered and sexual nature of EV with its own manifestations, motivations and causes (Bardall, 2011, 2015; Huber and Kammerud, 2016; Krook and Restrepo Sanin, 2016; Wanjiku Kihato, 2015), a theory and conceptual definition focusing on the form of SEV observed in Burundi has not yet been established. Gender implications of EV remain on the margin of policy and research, largely because it is situated in a disciplinary grey zone between conflict and democracy studies. Elections often lead to a return to wartime conditions, hinders the political participation of women and shapes the economic, social and political life of their countries (Bardall, 2011, p.3). There is still a gap in research when it comes to variations of causes of SEV against women, which hinders effective prevention and policy on solutions. Further, research is highly important in relation to the escalating humanitarian crisis and the horrendous, life-long effects for victims of sexual violence. In this in-depth single case study of SEV, I present three arguments. First, I make a theoretical contribution by suggesting that a novel definition is necessary in order to conceptualise SEV against women associated with the opposition. The fusion of previous research from different fields suggests that targeted SEV against women, similarly to wartime rape, can be used as an effective tool of political coercion at different stages of the electoral cycle. SEV serves the purpose to repress the opposition in proxy through the private sphere, which has different effects on victims, the opposition and the electoral process. Considering the overall electoral aim of this form of violence, it should be viewed as a distinct and gendered category of EV with its own dynamics. Second, I show empirically through exploring my definition in the case of Burundi that strategic rape is not exclusive to wartime

7 contexts. Women are targeted in specific ways as women, while primarily in their social and private role as wives or daughters, but targeted for their association to a man in the public sphere. Under sociocultural conditions where sexual violence and rape is highly stigmatised, SEV becomes an effective silent weapon of repression. Thirdly, I suggest that if SEV is indeed different from general EV and SGBV, this may have implications for how humanitarian action is currently conducted. Focusing on the overlap between the spheres of private and public in SEV, I offer a contribution to the wider debate on the aim of humanitarian action. The study will proceed as follows. The second chapter reviews the existing literature on EV and will make a novel contribution by introducing the concept of SEV. The third chapter describes the research design and how the empirical analysis will be carried out. The fourth chapter presents secondary empirical data from Burundi on SGBV, general EV and what has been defined as SEV. The fifth chapter consists of the analysis, contrasting the different forms of violence, and develops the arguments outlined above. In the sixth chapter, the study concludes that strategic rape is not something exclusive to wartime context. Under sociocultural conditions of high stigmatisation, sexual violence can be used as silent weapon of repression against the opposition. Reducing SEV to general SGBV is problematic for both policy-making and humanitarian action, as they require diverse action.

8 2. THEORY

This chapter presents previous research on EV. The examination leads to the argument that in order to shed light on the distinct category of SEV, one needs to combine EV theory with theories on SGBV. The chapter will also introduce the novel definition of SEV.

2.1 Previous research on electoral violence EV has no consistent definition in the social sciences, but efforts have been made to disaggregate the concept into classifications according to timing, the actors involved, and the intensity, motives and causes for violence (Bardall, 2015, p.4). One definition is brought forward by Fischer, who defined electoral violence as “any random or organised act or threat to intimidate, physically harm, blackmail, or abuse a political stakeholder in seeking to determine, delay or otherwise influence an electoral process” (Fisher, 2002, p.3). In many instances, EV is part of the general conflict dynamics between actors. Given the unique features of EV, it requires special attention both from scholars and policy-makers (Höglund, 2009, p.413). Rather than asking if electoral periods increase the risk of political violence, previous scholars have approached EV as a sub-category of political violence that is intimately linked to the electoral contest. EV is unique in timeframe, motive, target and assumed instrumentality and trajectories following an electoral result. The underlying assumption is that violence would not have occurred, or at least manifested itself differently, in the absence of electoral contest. Violence is used to affect the electoral process and its outcome, rather than trying to influence politics more broadly (Fjelde and Höglund, 2016, p.8). However, establishing an appropriate timeframe is difficult. Pre- and post-EV might be intertwined with the general dynamics of conflict, which weakens the link to the electoral process. However, researching EV as a dynamic of its own might be important for democratic outcomes and consolidation of democratic regimes, as violence hinders participation, undermines legitimacy of institutions and polarises intergroup relations (Höglund, 2009, see Appendix A for Höglund’s summary table). The causes of EV are multifaceted and can be explained by structural factors related to underlying power structures prevalent in transitional democracies, such as informal patronage systems, poor governance, exclusionary politics, socioeconomic uncertainty in states where power lies among elites, or electoral factors, democratic failure or election fraud, manipulated

9 institutions or the rules governing the process (Adolfo et al., 2012, p.1). Further, Hafner- Burton et al. (2014) have argued that incumbents are more likely to use violence when they have information about their unpopularity, which leads them to think that elections will decrease their power and the executive has few institutionalised constraints on their decision- making power (Hafner-Burton et al., 2014, p.149). However, this theory does not explain what form of EV incumbents turns to and the reason behind its sexual manifestation. Examining violence surrounding the election of 2007 in Kenya, Murunga (2011) suggests a relationship between abuse of electoral processes and the eruption of violence: pre- EV can have the consequence of increasing the probability of post-EV. If the legitimacy of the outcome is questionable and electoral choice is undermined through manipulation or abuse, it invites alternative means of achieving legitimate government. While protests can be seen as an alternative means of achieving legitimate government if the ballot box fails, they are harder to discipline and control than electoral results. This leads to an escalation of violence which further reveals the inability or failure of the political class to take charge and control its constituency (Murunga, 2011, pp.7–8). Some scholars have highlighted that the discussion on EV as a result of politicised ethnic identities, class and local socio-economic dynamics ignores how gender cuts across these categories and produces or shapes conflict. Wanjiku Kihato (2015) has argued that modes of violence cannot be explained by ethnic or political identities alone. If gender is ignored, we cannot fully understand why rape becomes part of the machinery of violence: gender cannot be disentangled from ethnic and political identity. The context of the 2007 elections in Kenya suggests that the ‘battle for political power was also a battle of masculinities’. Acts of SGBV became intensely entwined with ethno-political performances to annihilate opposing groups. The stake of losing a national election was closely linked to the loss of an ethnic machismo or manhood (Wanjiku Kihato, 2015, p.23). Bardall (2011) has argued for a gendered subset of EV, with distinct forms, causes, impacts of and solutions to violence. Previous definitions or frameworks of EV are male- oriented and non-gendered. They ignore familial or social intimidation and pressure that plays out in the private sphere. Drawing on research in domestic violence and feminist security studies, the component of sexual violence is introduced, including “politically motivated rape as a tool of terror and intimidation, marital rape as a tool of repression, sexual harassment, assault and abuse with the objective of controlling, intimidating, humiliating and disenfranchising” (Bardall, 2011, p.6).

10 Bardall briefly mentions that “historically, politically prominent women often became victims of election violence through their associations rather than their actions” (Bardall, 2011, p.12). Women have been politically visible as wives, mothers and daughters of political personalities, which has then made them targets for “political opponents seeking to intimidate or disrupt electoral proceedings” (Bardall, 2011, p.12). The intersection between EV and SGBV is acknowledged in Bardall’s latter work. EV becomes a form of SGBV when it is motivated by the aim to limit or deny electoral participation on the basis of sex or gender: “women’s electoral participation may be targeted for violence (…) to maintain patriarchal control of the political sphere” (Bardall, 2015, p.5). The International Foundation for Electoral Systems (IFES) offers a clear definition covering this concept of violence. Violence Against Women In Elections (VAWIE) is defined as the “harm or threat of harm committed against women with the intent and/or impact of interfering with their free and equal participation in the electoral process during the electoral period” (Huber and Kammerud, 2016, p.5). While covering the intent to interfere with women’s electoral participation through violent means, it does not cover cases of politically motivated violence against women where the goal is to interfere with the political participation of the men they are associated to. When Krook and Restrepo Sanin (2016) made the distinction between VAWIE and Violence Against Women in Politics (VAWIP), they highlighted some important insights on the relationship between the private and public sphere. While VAWIE includes acts directed at women as candidates, voters or activists, the subset VAWIP concerns violence perpetrated against female politicians with the purpose of leading them to withdraw from politics. In effect, the latter is directed against all women. Further, VAWIP often relies on gendered scripts, focusing on women’s bodies and traditional social role, questioning the belonging of women in the political sphere. Feminist theory makes an important contribution in highlighting the association in many societies between men and the public sphere of politics and economy while women are associated with the private sphere of home and the family. This divide “underlies many classical theoretical works in political theory which are premised upon the exclusion of women” (Krook and Restrepo Sanin, 2016, p.137). VAWIP seeks to reinstate this traditional divide, connecting women to stereotyped social roles and the private sphere (Krook and Restrepo Sanin, 2016, p.137). Further, the continuities across different manifestations of VAWIP is connected to distinct cultures that “provide different ‘tools’ for disciplining women, giving meaning to particular actions, as well as conditioning the degree to which such behaviours are normalised” (Krook and Restrepo Sanin, 2016, pp.138–139).

11 To summarise, questions remain surrounding the motivations, causes and manifestations of EV. In order to shed light over the dynamics of sexual violence towards perceived opposition-related women, one needs to combine theory on EV with theories on SGBV Sexual violence directed at women associated to the opposition shares the public/private sphere dynamics highlighted by Krook and Sanin, but when a woman is targeted for association, interfering with her free and equal participation in the electoral process is not the overall goal. These women are targeted in specific ways as women, but the aim is connected to their political association to a man. The public and private spheres thus overlap and the motivations, causes and manifestations of perpetrated violence against these women cannot be explained by EV theory alone. The following section will make an effort to fuse findings from different research fields and develop a conceptual definition applicable to contexts such as Burundi.

2.2 Focusing on the dynamics of sexual violence In this section, I make a contribution to previous research by outlining an argument that addresses the motivations, causes and manifestations of SEV. SGBV has been part of conflicts in the world for centuries (Pruitt, 2012, p.300), but empirical variation has hindered mono-causal explanations for its prevalence (see Wood, 2014). Further, scholars have called attention to how wartime sexual violence relates to other conflict behaviours and how it is carried over into post-conflict situations, both by actors of war and society at large (Nordås, 2011). It has also been debated whether SGBV is different in war and peace. The continuum thesis argues that conflict-related sexual violence is part of a continuum. The same gender relations that drive sexual violence during peace drive it during war. Patterns of sexual violence in peace and war might differ in degree, but not in kind (Cockburn, 2004). Further, the masculine conception of peace does not address the endemic conditions that produce violence and instability. The absence of a gendered understanding of the state and the nature of violence has contributed to the proliferation of SGBV even in areas considered peaceful and safe (Daley, 2008, p.230). The United Nations Fund for Population Activities use a definition that highlights rape as the most common form of sexual violence, but also includes “sexual threats, exploitation, humiliation, assaults, molestation, domestic violence, incest, involuntary prostitution (sexual bartering), torture, insertion of objects into genital openings and attempted rape” (UNFPA, 1999, p.36).

12 Similarly to EV, SGBV thus manifests itself in multiple ways. It also serves a variety of functions: an integral part of warfare; as an element of male-to-male communication or a tool of symbolic humiliation of one’s opponent; as a reaffirmation of one’s own masculinity; as a tool with which to destroy the culture of one’s opponent; and as an outcome of misogyny (Seifert described by Meger, 2010, p.122). According to Cohen et al (2013), wartime rape does not need to be ordered to occur on a massive scale. It is often not an intentional strategy of war, but more frequently tolerated than ordered. Further, states are more likely to be reported as perpetrators than other groups, which is considered puzzling as states are generally better trained and have better resources than rebel groups. Both factors are believed to reduce the likelihood of sexual exploitation of non-combatants, at least in the case of opportunistic rape. Evidence suggests that states frequently use sexual violence as a tool of torture against detainees as well as during operations against civilians (Cohen et al., 2013, p.4). In their cross-national analysis, Butler et al. (2007) argue that sexual violence committed by government forces and security police is more likely to be attributed to the selfish motivations of agents, under conditions of conflict, system-wide slack bureaucratic control and contexts of information and organisational constraints. Sexual violence may be its own reward for agents who can misuse their power and authority attached to their public office. The relationship between principal and agent is thus critical for understanding the prevalence (Butler et al., 2007, pp.669–670). Feminist theory argues that wartime rape is a manifestation of structural patriarchy. The chaos of wartime milieu encourages sexually violent behaviour (Gottschall, 2004, p.130). However, a theory of patriarchal culture fails to address the variation of the use of sexual violence by groups in the same context. While devaluation of women may be a core underlying cause for sexual abuse of women, the general idea of patriarchy is too broad to account for the variation in observations by several scholars (Wood, 2014, p.462; Cohen et al., 2013). The cultural pathology theory makes more of a cultural psychoanalytical claim: the history of nations and developmental factors are conspired to cause its men to descend to barbarism. A military culture fosters hostile attitudes towards women that culminate in the entitlement to rape. Sociocultural factors thus contribute to the frequencies of rape. While culture can explain some cases, it cannot explain the phenomenon as a whole. Massively different socialisation processes still seem to lead to rape (Gottschall, 2004, p.130). Strategic rape theory claims that rape is, similarly to other weapons of war, a tactic executed to spread terror, diminishing the resistance of civilians, and demoralising,

13 humiliating, and emasculating the enemy soldiers for not protecting their women. Rape splits familiar atoms of which societies are composed, causing unwanted pregnancy, shame, and abandonment by the family and community. It should be regarded as both a physical and psychological weapon (Gottschall, 2004, p.131). While strategic rape is carried out for the pursuit of group objectives, opportunistic rape is rape carried out for private reasons, such as sexual gratification or personal revenge. In light of these distinctions, it is likely that the pattern of rape by many groups that engage in frequent rape is neither strategic, nor opportunistic, but can rather be explained by theories on group cohesion (Cohen et al., 2013). Wood (2009, 2006) has argued for this view, claiming that the most plausible explanation for sexual violence is not the nature of conflict itself, but rather the internal hierarchy, discipline and ideology within a group. As rape occurs in varying proportions across conflicts in relation to other forms of violence, something else is needed to explain why sexual violence is chosen as a weapon of war, as well as when it is not used relative to other patterns of violence. In her latter work, Wood (2014) critiques the classical theories advanced to explain conflict-related sexual violence for not accounting for variation in context, prevalence and perpetrating groups. The militarised masculinity approach, highlighted by Goldstein, which claims that societies develop or draw on institutions and norms that include a highly militarised masculinity based on a sharp distinction between genders, cannot explain the absence of sexual violence in contexts where groups do not commit the crime (Wood, 2014, p.462). Similarly, the substitution argument, in which rape substitutes for sex under other conditions, cannot explain contexts where particular groups are targeted or why forces with access to sex perpetrate sexual crimes. Analysing the case of the Democratic republic of Congo, Meger (2010) found that the use of rape as a weapon must be understood in relation to both social constructs of masculinity and the politics of exploitation that have shaped history. The prevalence of rape was best understood from two dimensions: the individual motivation of combatants and the structural influences that are perpetuated in conflicts. Individual motivation for the use of sexual violence was largely informed by constructions of masculinity (Meger, 2010). One can question whether sexual violence and rape serves the same function in war and in contexts of electoral conflict. Little research has been done in these in-between contexts. Thomas et al. (2013) examined the relation between political transitions and SGBV, arguing that it is critical to recognise SGBV as bound to systemic gendered inequality. The feminist understanding is that the root cause for SGBV is not political turbulence, but patriarchy. Looking at the cases of South Africa, Zimbabwe and Kenya, political transition seemed to be

14 part of long histories of violence and each country had inherited legacies left by brutal regimes of colonial oppression and racial domination (Thomas et al., 2013, p.523). The scholars linked transition and its associated political and social upheaval to the increasing incidence of SGBV. In Zimbabwe, some sexual violence was seen as ‘message crimes’, sending a message to the opposition forces what would happen to an entire population if fighting continues. In Kenya, there was a clear upsurge of cases of women who had suffered sexual crimes after the transition, and gang rape surfaced during the political crisis (Thomas et al., 2013, p.524).

2.3 Towards conceptual clarity Scholars have encouraged more research on the different manifestations, causes and purposes of EV (Fjelde and Höglund, 2016, p.10). An examination of the underlying causes of sexual violence within electoral timeframes thus makes a contribution to gather more understanding of the complex connections between wartime sexual violence, EV and the contingencies of conflict legacies. Drawing on research and definitions described above, it becomes clear that some of the acts reported in Burundi relating to the election of 2015 do not fit into the existing definitions, even when gender is taken into account. Therefore, in this study, sexual electoral violence (SEV) is defined as the following:

Sexual electoral violence is the random or organized act or threat to physically harm, abuse or assault a perceived political stakeholder sexually, occurring within the timeframe of elections, with the purpose to repress the adversary or with the aim of influencing the electoral process, outcome or legitimacy.

The definition is intentionally gender-sensitive: men can also become victims of SGBV and SEV. However, this study will focus exclusively on the targeting of women. Further, considering the overall aim, this form of violence should be viewed as a distinct and gendered category of EV rather than VAWIE. It is also distinct from general SGBV due to its political dimension. The fusion of previous research suggests that similarly to wartime rape, SEV can be used as an effective tool of political coercion within the timeframe of elections under circumstances where sexual violence and rape is highly stigmatised, which makes it an effective silent weapon of repression. At various stages of the electoral cycle, SEV serves the

15 purpose to repress the opposition in proxy through the private sphere, which has different effects on victims, opposition and the electoral process. Women are targeted in specific ways as women, while primarily in their social and private role as wives or daughters, but targeted for their association to a man in the public sphere. Being a woman determines the form of violence, the manifestation of sexual violence; the association to a political opponent explains the cause or motivation for EV.

16 3. RESEARCH DESIGN

This chapter describes the research design used for substantiating the theoretical argument with empirical material. This study conducts an in-depth single case study of SEV against women in Burundi with the purpose of theory-building (see George and Bennett, 2005, pp.111–115).

3.1 Research design A single case study has apparent limitations (King et al., 1994) but if understood as a particular way of defining cases, not a way of analysing cases or a way of modelling causal relations, a case study can still be an efficient approach. Case studies enjoy a natural advantage in research of an exploratory nature. In-depth case study research allows for the generation of hypothesis, insights that might go missing in cross-unit research (Gerring, 2004). Further, a case-based research strategy is distinguished by the fact that it examines the contemporary phenomenon in its real-life contexts, which is especially useful when the line between phenomenon and context is thin. A case study approach always has to deal with the problem of too many variables, as context is part of the study (Yin, 1981, p.59). Notwithstanding these obstacles, the inductive approach of this study allows for unique insights. While this study culminates in context-specific knowledge, it still paves way for theoretical and conceptual development. The conclusions are most likely to apply in post- conflict states that have a similar history of wartime rape and EV.

3.2 Case selection How does one pic the topic for analysis? While case selection is important for the value of research, “…there is no such thing as a logical method of having new ideas… discovery contains ‘an irrational element’, or a ‘creative intuition’.” (Popper in King et al., 1994, p.14). In this study, case selection is somewhat intertwined with the idea of developing new ideas and theory: the study departs from the empirical observations of SEV in Burundi, linking them to the conceptual definition developed in the theoretical chapter. The variation in the form of violence (dependent variables) is explained by the causes and motivations behind violence (independent variable). While previous frameworks for categorising EV separate

17 causes into enabling conditions and triggering factors (see Höglund, 2009, p.423), this study categorises them more broadly as one. The case also has a history of wartime rape, high levels of SGBV and general EV, allowing me to contrast the three phenomena. A sufficient explanation is not being built, rather, a relatively parsimonious mechanism is uncovered that is suggested to contribute to the outcome. Further, the case is selected as a ‘deviant case’, one that has outcomes that are caused by mechanisms known from other research but have been overlooked in relation to the case (Beach and Pedersen, 2013, p.154). Without compromising validity, the initial theory is inspired by the deviant case, then tested again against additional evidence, which was not used to derive the theory (George and Bennett, 2005, pp.111–112). It is always difficult to devise a program of falsification the first time a new theory is proposed. Path-breaking research is by definition exploratory in nature (Gerring, 2004, p.349) The ambition is not to prove that a theory is correct, but rather to prove that is has utility in providing the best possible explanation. Social scientific explanations are always “case specific and cannot be detached from the particular case” (Humphreys in Beach and Pedersen, 2013, p.13).

3.3 Structure of empirical description and analysis The empirical chapter is divided in three parts: general sexual violence, EV and SEV. This distinction is made for analytical clarity: in order to separate SEV as a sub-category of EV and distinct phenomenon, it needs to be contrasted with the other categories. However, it is acknowledged that the forms of violence are sometimes intertwined and entangled in reality. Each part will include the profiles of victims and perpetrators, the manifestation of violence and the motivation or cause of violence. In order to answer the research question and explore the theoretical argument, the analysis consists of a qualitative assessment of the following sub-questions:

§ Who are identified as perpetrators and victims of the different forms of violence? § How are the different forms of violence manifested? § How are victims or perpetrators linked to underlying causes? § Are SEV events associated with political threats or propaganda? § Does the government deny its association to events of SEV? § How do victims of SEV describe the crimes, possibilities to report and overall response from society after being victimised?

18

The questions can be seen as an operationalization of SEV, which guide the analysis of the empirical background and analysis. Direct indicators supporting the theoretical argument are the explicit statements from ruling-party loyal groups that include the motivation for sexual violence, political propaganda or political threats involving women associated to the opposition and observed involvement of ruling-party loyal groups in events if SEV. Indirect indicators are mass-flight of perceived opposition leaving women unprotected, as well as government responses to accusations.

3.4 Material and delimitations The material consists of a variety of data from organisations, scholarly articles and daily newspapers. The material was chosen departing from the checklist of source criticism and according to standards of relevance, validity and triangulation described by Höglund and Öberg (2011, p.188). As SEV has not been studied previously, the study had to rely on secondary sources, possibly biased or access-dependent organisations. Further, it is possible that some women have reported sexual violence twice to organisations when translocated (HRW, 2016a) and underreporting remains a challenge to any research on the topic. However, data from different sources was compared and triangulated, strengthening the reliability of the study. Further, the volume is not necessary for exploring the theoretical argument. The research validity is considered high as the operationalization, guiding questions and chosen indicators capture the phenomena studied. While a high rate of general SGBV might complicate the unravelling of specifics about SEV, the inter-connections between the phenomena studied are seen as contributing to the exploration.

19 4. OBSERVATIONS OF VIOLENCE IN BURUNDI

This chapter describes SGBV, general EV and what has been defined as SEV, providing an overview of the broader social context in which SEV is situated. Descriptions of individual incidents play an important part for the analysis in chapter five, providing foundation for the theoretical argument that SEV functions as a silent weapon of repression under circumstances where sexual violence is highly stigmatised.

4.1 SGBV and the legacy of civil war

“Press charges? I told the Chef de Quartier and the local radio – what else could I have done? Besides, what could I have done against a soldier?”

“NS”, Bujumbura, victim of rape in 2001. (Quoted in Amnesty International, 2007, p.7)

Burundian society is highly patriarchal: sexual violence in Burundi takes place in a context of protracted crisis that stems from the civil war. Throughout history, there has been widespread underreporting of SGBV because of social stigmatisation, exclusion, fear of reprisal, impunity, lack of confidence in the authorities to investigate and the lack of funds for travel costs to visit medical or counselling centres (Daley, 2008, p.125; Pézard and De Tessières, 2009, p.76; HRW, 2016a; Amnesty International, 2007). During the war and its aftermath, rape was deliberately used by armed groups and government forces as a means of “denigrating, humiliating and implanting fear among women, their communities and the civilian population” (Amnesty International, 2007, p.7). Statistics “represents only the tip of the iceberg” (Amnesty International, 2007, p.5). Motivations varied and the use of sexual violence has been connected to militarist masculinity, strategic warfare, and acts of reprisal or tit-for-tat, as well as simple opportunistic brutality. In relation to the history of genocide, the sexual violence has also been connected to the sanctity of ethnicity, religion and the nation (Daley, 2008, pp.125–126). Sexual violence served multiple functions: to destroy or seriously impair the opposition, to instil fear into the local people and to emphasise the military capability to external observers (Daley, 2008, p.230). As a manifestation of the intersection of racial,

20 ethnic and gendered hierarchies, certain women differentiated by race and ethnicity who where seen as lower down in the human and feminine hierarchy were exposed to forms of violence that “defile their gender and person” (Daley, 2008, p.124). Women of every age were at risk in the context of war, in particular those living in camps and women from wealthier households (Dijkman et al., 2014, p.21). The geography of reported rape during the civil war and its aftermath suggests a strong correlation with areas of intense military activities, greater sensitisation to the criminality of sexual violence and available assistance to victims. When the army and FNL started fighting again in 2008, it triggered sexual violence in the fighting-regions, declining with the cessation of armed conflict (Daley, 2008, pp.125–126). While all sides committed rape during the conflict, CNDD-FDD committed more sexual violence throughout the war and afterwards. After CNDD-FDD’s democratic accession to power, between 2005 and 2011, reports of sexual violence acts committed by CNDD-FDD were more numerous than those committed by oppositional FNL. Between 2004 and 2007, the FNL was not associated with reports of sexual violence (Muvumba Sellström, 2015, pp.152–153). While there was a code within the CNDD-FDD prohibiting sexual violence, the instrument was flawed and the authorities openly flouted the use of penalties. Authorities minimised punishment for perpetrators because they were complicit in sexual violence. The opportunity to assume power as a political party proved critical to changing the practices regarding sexual violence. After settlement, the CNDD-FDD was more serious about the prohibition, but a culture of impunity had already been established (Muvumba Sellström, 2015, pp.175–181). According to evidence from NGOs compiled by Daley (2014), SGBV was even more widespread in the immediate post-war period as reported cases of rape rose dramatically after the peace-agreement in 2003. Most perpetrators were identified as community members know to the victims. Ligue Iteka saw a rise from 983 cases in 2003 to 1791 cases in 2005; Nturengaho reported 93 cases in 2003 and 446 in 2004; Seruka Centre received 1119 cases between September 2003 and 2004. The organisations attributed the increase to the demobilisation of young men from the army and rebel forces (Daley, 2014, p.70). Over a five- year period, extremist actors were neutralised and a political culture emerged that “favoured de-escalation and compromise; at the same time, large numbers of rebels where integrated into the army, up to the highest level of command” (Uvin, 2010, p.162). The sociocultural order and tradition of male superiority has facilitated the discrimination and repression of women, including sexual violence. The subordination of women is institutionalised through Burundian law: for example, men are considered the head

21 of the family and women cannot inherit land. Women are often more closely aligned with property: rape is seen as vandalising a man’s property, but without reprisal or legal consequences (Hoofnagle and Rothe, 2010, p.184). Subordination is accepted and tolerated, consciously or not, by both men and women, particularly in rural areas (Pézard and De Tessières, 2009, p.76). While recently reported incidents of SGBV seems to be more of a political character, HRW also that some attacks were related to land. Many women who were attacked in their homes had returned between 2010 and 2012 to Burundi from refuge in Tanzania, finding that neighbours or family members had occupied their homes (HRW, 2016a). According to Dijkman et al., (2014) experts have “regularly indicated that victims of sexual violence have become younger throughout the post-conflict period” (Dijkman et al., 2014, p.11). It has been suggested that this trend reflects a breakdown of societal norms as well as the fact that during the war, rape was often committed against unprotected adult women who went into the field or forest in search for firewood and water (Dijkman et al., 2014, p.11). Data from victim-assistance centres and experts suggested that “today’s sexual violence perpetrators are predominantly family members or in-laws, male domestiques, and – very often – neighbours” (Dijkman et al., 2014, p.12). Further, poor middle-aged women in urban areas who had finished primary schooling are more at risk to suffer from general SGBV. When the husband was identified as the perpetrator, it was related to domestic and intra- marital violence. Neighbours or community members, often demobilised ex-combatants and military who were still in possession of arms, often perpetrated sexual violence within the household they were employed. The end of the civil war reduced the number of rapes by strangers, as a result of reduced population mobility (Dijkman et al., 2014, pp.12–13).

4.2 Electoral violence in Burundi and the crisis of 2015 The electoral crisis of 2015 started on 25 April when president Nkurunziza announced to run for an unconstitutional third term in office in the upcoming election in June (for key events, see Appendix B). Burundians took to the streets in the capital Bujumbura against the backdrop of “sustained militarism, political and economic repression, intimidation and exclusion, and widespread fear” (Daley and Popplewell, 2016, p.648). Between April 2015 and April 2017, UN figures indicate that over 400 000 people have fled Burundi because of the political crisis (OHCHR, 2017). Similarly to the previous election, the state intensified authoritarian practices in the election of 2015. On 13 May 2015, Nkurunziza survived a coup-

22 attempt whilst attending a meeting of the East African Community in Tanzania. Violence persisted after elections, with tit-for-tat killings of the perceived opposition, intimidation and imprisonment of dissidents (Daley and Popplewell, 2016, pp.648–649). A culture of impunity was consolidated along the ruling-party’s tightening grip over state institutions, allowing for law-reformation, extra-juridical killings, low-intensity violence and the closing of channels of free speech (Daley and Popplewell, 2016, p.655). The reported perpetrators of violence belonged to security forces, intelligence agents, Imbonerakure, police and other groups loyal to the ruling-party, but armed oppositional groups have also attacked security forces and ruling-party members, including the police and Imbonerakure (HRW, 2017). Further, the opposition is organised around one civilian platform and a number of armed groups. It remains divided and its unification illusory. It has been difficult to pinpoint the responsibility for acts of violence as several oppositional groups have claimed that they carried out particular attacks, mostly against the security forces (International Crisis Group, 2016, pp.17–19) The United Nations Independent Investigation on Burundi (UNIIB) released a report in September 2016, finding that gross human rights violations had and are taking place in Burundi, committed primarily by state agents and linked groups in a systematic and patterned way. Violence was largely motivated by efforts to extract information about or directly hurt perceived opposition, spreading fear and reprisal attacks (UNIIB, 2016). The initial street protests were held in a non-violent fashion. After the attempted coup, the government started associating all dissidents with the military coup leaders and the opposition. The attack on four military bases in the capital province on the 11 December escalated violence against civilians and expanded targeting to state officials, intelligence service agents (SNR) and senior military officers suspected of dissidence (UNIIB, 2016, pp.5–6). The UNIIB reported that more than a thousand had been killed, thousands repeatedly tortured, unknown number of women had been a victim of various forms of sexual violence, hundreds have disappeared and thousands were illegally detained. Further, allegations of mass burials of executed people were widely reported. Human rights activists as well as their relatives were reported to have been primary targets of violence (UNIIB, 2016, pp.7–8). The forms of violence varied: UNIIB reported not only about killings and violence against protestors in the streets, but also on forms of torture and inhuman and degrading treatment. Reported incidents of torture included attaching weights and injections to genitals, crushing limes, detention in a closed container, forcing families to stay with dead relatives or during gang-rapes and humiliating speech including an ethnic dimension (UNIIB, 2016, p.10).

23 4.3 Sexual electoral violence surrounding the electoral crisis of 2015

“I was held by the arms and legs. [An attacker] said: ‘Let’s kill her, she is an [opposition National Liberation Forces] FNL wife’ as they raped me.”

Victim of targeted rape in Bujumbura, October 2015. (Quoted in HRW, 2016a)

After the 2015 elections in Burundi, organisations such as Human Rights Watch (HRW), Amnesty International, the Office of the United Nations High Commissioner for Human Rights (UNHCRC) and United Nations High Commissioner for Refugees (UNHCR), reported about an escalation of sexual violence, suggesting that groups loyal to the ruling-party were raping women suspected of being connected to the opposition. Several journalists highlighted the horrendous experiences and consequences for victims (HRW, 2016a; McNeish, 2016; Sanghani, 2016; Steers, 2016; United Nations News Service, 2017). In July 2016, HRW released a report covering 323 cases of sexual violence against women from May through September 2015, numbers documented by the UNHCR. The pattern of sexual crimes appeared to have been aimed at family members of perceived government opponents, showing covariance with general EV against male family members, sometimes members of the opposition, in the form of killings or abduction. The victims of sexual violence often identified the perpetrators as members of Imbonerakure, the militant youth wing of the CNDD-FDD, or policemen, security forces and military soldiers either dressed in uniform or civilian clothes. UNHCR documented that of all reported cases between June and October, “16 where allegedly perpetrated by the police and 177 by other security forces or the Imbonerakure” (HRW, 2016a) The HRW conducted more than 70 interviews in Tanzanian Nduta refugee camp with victims of sexual violence, where “dozens of women said they were raped in or close to their homes” (HRW, 2016a). Men armed with guns, sticks or knives raped or sexually assaulted women during night time attacks or at the Burundi-Tanzanian border as they tried to flee the escalating political crisis. More than 10 of the interviewed women said that local Imbonerakure had harassed them before the announcement of the third term, although it got worse afterwards. 38 of the women recognised the perpetrators, while in 23 cases the victim did not recognize the perpetrator. However, some women in the latter group stated that the men were dressed in ruling-party t-shirts or police uniforms, common among the

24 Imbonerakure (HRW, 2016a). Women also told HRW that if male family members where absent, the perpetrators would demand his whereabouts. In 9 cases, women stated that the men had fled before the rape took place; in 3 cases rape took place weeks after a male family member had been abducted or killed. The wife of an FNL-member stated that her husband had been verbally harassed multiple times before he was abducted. However, the motive was not always clear-cut political. In one case, attackers took a mobile phone and in two other cases, they took money, but “robbery did not appear to be their main motivation” (HRW, 2016a). Further, victims linked some of the attacks by groups loyal to the ruling-party to local or personal disputes, grievance or ethnicity. The FNL-husband of another victim had reported local men to the police for buying sexual services from a prostitute, and they were arrested. The victim quoted the attacker saying “he [the husband] stopped us from using their prostitute, we we’re [having sex with you] instead” (HRW, 2016a). Sometimes ethnic slurs were made against Tutsi women during the attack. In 5 cases, raped women said they feared repeated attacks from Imbonerakure whom they claimed were in the Tanzanian camps to target people (HRW, 2016a). The UNIIB report confirmed documentation from HRW, claiming that they had “obtained credible information indicating that many Burundian women and girls related to males who opposed the third term, or were perceived as political dissidents, became the targets of physical and sexual violence by elements of the security forces” (UNIIB, 2016, p.10). Local health workers in Bujumbura believe that there was a surge in sexual assaults in the end of 2016 connected to the increase of military and police operations in opposition neighbourhoods. A senior advocate at Refugees International noted that women became more vulnerable as men disappeared or fled the opposition neighbourhoods. Sources at a UN agency funding a community heath centre reported that Imboneakure demanded bribes near health centres, asking why the victims of rape were seeking medical care, hindering women from adequate care (Steers, 2016). They also stated that they believed the actual numbers of rape to be much higher: “from our work here we know that when you have one case in the centre, you have 20 in the neighbourhood” (Quoted in Steers, 2016). Rape has not been the only manifestation of SEV and victims have not exclusively been women (for cases of male victims, see HRW, 2016b; UNIIB, 2016). In the UNIIB report, a woman reported to have been sexually mutilated by the Imbonerakure in August 2015 in a search: “They put their hands inside my vagina until the uterus came out. I was left alone bleeding, screaming. The neighbours came out and they tried to put my uterus back in place” (UNIIB, 2016, p.10). A witness stated that in April 2015, armed men dressed in military

25 uniforms had gone in to her neighbour’s house and the following day, the woman was found dead with wood inserted in her vagina. Both incidents happened in the province of Cibitoke (UNIIB, 2016, pp.10–11). On the 1st of April 2017, about 200 members of the Imbonerakure gathered in a ruling party-office in Kirundo province, northern Burundi. They were caught on video singing “Impregnate the opposition so they give birth to Imbonerakure… There are lots of girls. Impregnate them, Imbonerakure!” (Hartill, 2017). Local authorities did not attempt to stop them. Shortly after the video caught international attention, the government issued a statement saying it would open an investigation: “unfortunately, contrary to the ideals of the [ruling party], certain youth sang a song that doesn’t respond with the morals or the ideology of the party” (Hartill, 2017). Sources told Cape Times that such rape-songs had been sung since the election campaign for Nkurunziza’s third term started, while an exiled artist, Kaka Boney, claimed that it had circulated among the Imbonerakure since 2010 (Ikporr, 2017). Reports indicate that larger rallies have been organised by government officials or the ruling party. On the same day as the rally in Kirundo, around 2500 Imbonerakure members marched in northern province Kayanza, chanting similar songs, inciting rape and violence against opponents. Reports also suggest that this occurs on a weekly basis at Imbonerakure meetings in southern province of Makamba. Following the inauguration of a ruling party office in eastern province Ruyigi on the 8th of April, around 200 people began chanting a similar song inciting rape of opponents and to impregnate women, although party officials instructed them to stop. Dozens of Imbonerakure members reportedly marched in Gaharo of the southern province of Rutana the same day, chanting hostile slogans. The UN High Commissioner of Human Rights, Zeid Ra’as Al Hussein, expressed deep alarm at the “widespread pattern of rallies (…) where young men from the Imbonerakure militia repeatedly chant a call to impregnate or kill opponents” and described it as a “tip of the iceberg, brought to light only because it was captured on camera (…) The condemnation [by the ruling-party] is meaningless if, instead of putting a stop to such events, senior government officials continue to take part in such rallies” (OHCHR, 2017). Much like the diverse ruling party, Imbonerakure consists of different members of the Burundian society. The average age is between 25 and 35, and some are more educated and have gone to university, while others have fought in the civil war. The promise of development has made some join the group, who see the CNDD-FDD as a path to political and economic power or becoming involved in business. The political connection has somewhat provided an escape from poverty: the promise of cash for committing violence has

26 been reported in more rural areas. While estimates refer to the official membership of 5000, the armed number is estimated as be as little as 500, but is probably an underestimate as many identify with the group without being an official member (Buchanan, 2015). The government is still claiming that the Imbonerakure is simply a political group. The state police have been reported of stopping Burundians from fleeing the country and the Imbonerakure are “at every corner” (Sanghani, 2016), hindering people from crossing the border. One victim described the brutality when she was raped in a cell by a policeman after having tried to make it to the Tanzanian border with her family and was arrested by Imbonerarkure, local official and policemen. “I was shouting and struggling, but he did what he wanted. (…) He only raped me once because there was blood and urine coming out, and it hasn’t stopped since.” Many victims of rape at checkpoints within Burundi and at border-crossings described how rape was used as punishment for the accusation of being associated with the wrong political party (McNeish, 2016). Two victims described how police and Imbonerakure beat and undressed them while saying “why are you fighting the president?” (Steers, 2016). This chapter has presented empirical data on different forms of violence in Burundi. The next chapter will contrast them, arguing that SEV can be used as an effective silent weapon of repression against the opposition.

27 5. CONTRASTING SEXUAL ELECTORAL VIOLENCE

This chapter presents the analysis, contrasting SEV with general EV and SGBV. I will argue that SEV, similarly to wartime rape, can be used as an effective tool of political coercion within the timeframe of elections under circumstances where sexual violence and rape is highly stigmatised, which makes it an effective silent weapon of repression. Further, alternative explanations and methodological problems will be elaborated on. The chapter will end with a discussion on the implications for humanitarian action.

5.1 The dynamics of violence In Burundi, the different forms of violence show a variation in perpetrator, victim, manifestation and causes and motivations. In the civil war, all women where at risk of becoming victims of sexual violence, especially women living in camps and women of wealthier backgrounds. The perpetrators were mainly from the CNDD-FDD, but all fighting groups committed sexual crimes in the form of rape or gang-rape. Post-conflict SGBV mostly manifested itself as domestic abuse and rape, occurring within the family or community, targeting younger women than before. The perpetrators were mostly identified as demobilised ex-combatants, family or community members. General EV was committed by ruling-party loyal groups and occasionally oppositional forces, mainly targeting perceived opposition groups or individuals suspected of dissidence. Women perceived to be associated with the opposition became victims of the distinct form of SEV, targeted exclusively by groups loyal to the ruling-party.

5.2 The dissimilarity of sexual electoral violence The theoretical argument of this study holds that SEV serves the purpose to repress opposition through the private sphere, which has different effects on victims, opposition and the electoral process. Women are targeted in specific ways as women, but also for their political association to a man. Being a woman determines the form and manifestation of targeted violence; the association to a political opponent explains the cause or motivation for EV. Contrasting the different forms of violence in Burundi reveals that SEV showed a distinct dynamics. Several indicators of SEV were supported empirically. Descriptions of

28 explicit statements during rapes about the political affiliation of male family members, sexual threats against opposition-associated women through propaganda rape-chants, the involvement of ruling-party loyal groups in events of SEV in the homes of the victims, serve as direct indicators SEV. The pattern of male family-members fleeing political repression, leaving women unprotected, and the government’s denial of accusations serve as indirect indicators of SEV. The indicators simultaneously confirm the distinction between general EV and SEV. Table 1 summarises the findings. As suggested previously, the patriarchal sociocultural context facilitated conditions where SEV could be used in a similar fashion to wartime rape.

Table 1. Summary of findings

Form of violence Perpetrators Victims Manifestation Causes and motivations

Conflict SGBV CNDD-FDD All women, especially Rape Impair oppositional forces Other armed groups those living in camps Gang-rape Strategic warfare (emphasize and wealthier military capacity or humiliate or denigrate opposition) Reprisal or tit-for-tat Opportunistic brutality Culture of impunity Access to arms

Post-conflict Demobilised ex- Young women Domestic abuse Patriarchal structures or repression SGBV combatants Rape legitimising violence Husbands, neighbours, Culture of impunity family and community Opportunistic brutality members Legacy of rape as strategy of war (demobilisation of combatants) Institutionalised subordination of women Access to arms

Electoral Imbonerakure Opposition politicians Assassination Electoral conflict on third term violence Police Perceived opposition Torture Patrimonial politics Security forces Military coup leaders Abduction Conflict cleavages Intelligence agents and associated Violence as a legitimate political Ruling-party loyal Protestors tool groups Civilians Culture of impunity Oppositional groups Access to arms

Sexual electoral Imbonerakure Women associated Rape Causes or motivations of EV violence Police with the opposition Gang-rape Security forces Women with Sexual mutilation Stigmatisation Ruling-party loyal perceived association Sexual torture Patriarchal structures groups with opposition Women living in à Silent weapon of repression opposition-stronghold neighbourhoods Female refugees

29 While the pattern of sexual violence within the timeframe of the 2015 election shares similarities to the conflict era, the motivation behind sexual violence seems to be different: I suggest that electoral competition is a larger driving force than ethnicity as targeting is dependent on perceived political affiliation rather than ethnic identity. This might be a result of the change of electoral politics from ethnic parties to political parties, or simply the transformation of conflict. However, some victims mentioned ethnic slurs. The latest reports from Burundi indicate that ethnicity might become a dimension of conflict again, which might stem from a strategic choice to ease mobilisation or a backdrop to conflict dynamics. SEV was used to repress the opposition silently, as sexual violence generally does not lead to any legal consequences. Through the expansion and consolidation of ruling-party control over state institutions, the culture of impunity could prevail and rape could be used as a silent weapon in a highly stigmatised sociocultural context. Furthermore, is important to note that incidents of gang-rape and politically motivated rape increased dramatically during the last political crisis according to the reports. These sexual acts of violence are deeply embedded in the history of ethnic and political conflict manifested again in relation to the events of the 2015 elections.

5.3 Alternative explanations and critical reflection There are several possible alternative explanations to the motivations and causes for what I defined as SEV. One plausible explanation could be the contingency of historical factors and the legacy of strategic rape. The difference would lie not in the manifestation of rape, but rather the purpose. While used as a form of strategic warfare with an ethnic dimension during the civil war, in the electoral timeframe, it was more of a weapon to influence the process, outcome and legitimacy of elections. The elections certainly made Burundi return to wartime- similar conditions. Prescribing the current manifestation of sexual violence to either historical legacy or the electoral context becomes tricky. However, if one looks to the neighbouring Rwanda, it has shown no signs of targeted SEV to my knowledge, while sharing a similar history of ethnic conflict and wartime rape. Rather, Rwanda has made great accomplishments in closing gender gaps and reducing SGBV (Mbabazi, 2015). In Burundi, the capability of the ruling party to mobilise or allow individuals to perpetrate the specific form of SEV is probably enhanced or somewhat linked to the previous patterns of violent events in history, following the protracted nature of endemic genocidal violence against parts of the population. One can imagine that demobilised combatants that have committed rape during the civil war

30 and within communities in the post-conflict context do not refrain from committing sexual crimes within the timeframe of elections. There is a possibility that SEV is an opportunistic act that results from the unequal power-dynamics that emerges when armed male perpetrators enter the home of an opposition member with the aim to kill or abduct him and finds his unarmed wife or daughter. However, theories assuming opportunistic behaviour cannot explain why perpetrators enter the homes of opposition members in the first place, especially if they already have information about the absence of the male opposition member. While Thomas et al. (2013) argue that patriarchy is a driving force behind a chosen form of violence, it is inadequate alone to explain the case of Burundi and why groups loyal to the ruling-party are targeting opposition women specifically. Further, both the gender and nationalism thesis can provide an understanding to the specific nature of sexual violence in the context of genocide, but they do not explain its pervasiveness in Burundi in relation to the 2015 election. Another explanation could be the group-cohesion theory that has been argued by several scholars, such as the often-cited Wood. Multiple relationships and dynamics are at play when entering the private sphere of someone’s home with a political aim that is not realised. Sexual crimes could thus be connected to the hierarchy, discipline and ideology that members belong and adhere to. It could also be a form of sexual gratification, greed or personal revenge, when not able to fulfil one’s duty in finding the actual target. However, in the case of Burundi, it is impossible to explain the use of SEV by relying only on individual motivations or systemic influences. It is not only a product of conflict. I argue that the many reports about the Imbonerkure rape-chants are a strong indicator of a pattern of violence more similar to strategic rape-theory, but within a context of elections. Group-cohesion only accounts for the dynamics within the group and not the webs of relationships between conflicting parties, the influence from the sociocultural norms in society as a whole. Rape is not situated in an isolated context. The strong power-position of the ruling-party that saturates Burundian institutions and governing apparatus in relation to the weak and fragmented opposition cannot be ignored as a force of influence on the form of violence. The most probable explanation for SEV is that it is a silent weapon, effectively used within a context of stigmatisation, lack of legal repercussion and a culture of impunity in Burundi. Some questions remain unanswered. Victims of SEV could be more likely to report about the crimes of the ruling-party to humanitarian agencies than women victimised by other groups. However, I assume that they are less likely to report, as the level of trust for agencies should have gone down if perpetrators and victims are living side by side in camps. It is also

31 not clear if reporting affects the political competition and chances to run for office on behalf of both sides. Further, the comparison with Rwanda might be too simple: SEV might simply not be reported because of authoritarian politics. Further, the reports used in this study are submitted by agencies with a humanitarian agenda, but they are currently the only reliable source of information, as the Burundian state does not collect data in a systematic and impartial way.

5.4 Implications for policy and humanitarian action

“I became a child refugee to escape rape. It didn’t work.”

A young victim of gang-rape, Nduta refugee camp. (Quoted in Sanghani, 2016).

If SEV is indeed different from EV and SGBV more generally, this may have implications for how humanitarian action is currently conducted. Reducing SEV to general SGBV is problematic, as they require diverse action. While both categories of victims need access to medical care, victims of SEV can be assumed to be less likely to seek health care, to prosecute perpetrators and are in greater need of protection. The electoral dimension of sexual violence described in this study shows that the spheres of private and public overlap. This requires humanitarian agencies to rethink their strategies to reach and protect victims while upholding humanitarian principles. While armed conflict is violence that threatens the socio-political order, violence against women does not render societies ungovernable but is rather a component of existing systems of power (Thomas et al., 2013, p.521). The international discourse on sexual violence in conflict separates it from the continuum within which a culture of violence breeds: UN approaches have treated sexual violence in peacetime and conflict as separated intervention issues, where conflict rape requires international attention, while sexual violence during peacetime is easily perceived as the lesser evil. After conflict, SGBV is treated as a humanitarian issue to be taken care of by relief measures (Ayiera, 2010, p.15). Allocating resources towards victims of SEV when the perpetrators are connected to those in power becomes a choice between classic, Dunantist humanitarianism and its new, more political counterpart. While the widely accepted definition of humanitarianism as the “impartial,

32 independent and neutral provision of relief to those in immediate danger or harm” (Barnett, 2005, p.724) emerged in opposition to politics, it was recognised that it has political consequences “inextricably part of the political world” (Barnett, 2005, p.724). Today, it serves a new function in international politics as a project designed to transform the world rather than restricted to emergency relief. Yet the transformation also serves an ideological function, helping to reproduce the geopolitical order because it reduces pressures that might demand its transformation. While many liberal values have expanded the scope of humanitarianism, most activities have not challenged the Westphalian state-system (Barnett, 2005, p.733). After all, humanitarian agencies are dependent on access and the delegation of governance. Allocating aid to victims of sexual violence in Burundi without addressing the compliance of the ruling-party reproduces its legitimacy, but most importantly, it allows for violence to continue. Therefore, the case of Burundi intrinsically involves the division between the classic, Dunantist philosophy and the new interpretation of humanitarianism – a division at the heart of humanitarian action.

33 6. CONCLUSIONS

6.1 Main conclusions Previous scholars have explored the forms, causes, impacts and solutions to EV, but little attention has been paid to gendered forms and its sexual manifestation. Developing a theoretical argument from existing research, my study shows that the use of strategic rape is not something exclusive to wartime contexts. Electoral competition can fuel strategies previously only observed in wartime contexts. Further, this study highlights how reducing SEV to general SGBV ignores how the private and public sphere overlaps. This is problematic, as they require diverse action through policy and humanitarian action. This study identifies stigmatisation of sexual violence and the patriarchal sociocultural context of Burundi as the main causes behind SEV. Together with a culture of impunity it results in an effective silent weapon of repression against the opposition, motivating the specific sexual manifestation. The case of Burundi suggests that there is a preliminary value to conceptualising SEV as a subset of general EV, as they are similar in the purpose to repress the opposition with the aim of influencing the electoral process, outcome or legitimacy. The case explores how women are targeted in specific forms as women, primarily in their social and private role as wives or daughters, but targeted for their political association to a man. Contrasting different forms of violence in Burundi suggests that SEV, EV and SGBV varied in their manifestation, perpetrators, victims, and causes and motivation. SEV seemed to share more similarities to SGBV in the Burundian civil war than general EV, as they both functioned as a strategy. SEV targeted the opposition in proxy through sexual violence against (perceived) opposition-associated women, while EV targeted the (perceived) opposition and military coup-leaders directly in the form of assassination, torture or abduction. Further, protestors and civilians were targeted in different ways regardless of their political affiliation.

6.2 Future research It is acknowledged that this study has its obvious limitations. In-depth single case studies are hard to generalise and always need to be confirmed with broader studies. This study has provided an initial theorisation and exploration of SEV. Further research is needed to expand our knowledge about its function as well as the variation in contextual factors motivating political actors. Following my argument, we should expect SEV to be higher in localities of

34 high political escalation, where two or more parties could potentially win, than areas of low political contestation. This could be tested using a subnational quantitative or qualitative design. If correct, we would also expect SEV to be higher in countries with high political contestation compared to countries with low political contestation, which could be tested using a cross-national quantitative or qualitative design. Studying SEV over time, comparing data on its prevalence and nature in previous elections could account for variations in relation to policy changes and the legal framework on rape. If correct, changes in policy and the legal framework that counter impunity for sexual violence should decrease the motivations behind SEV. Furthermore, scholars should investigate the possible scope of the argument by looking at conditions such as if the general participation of women in politics matters for the prevalence of SEV, as well as the effect of economic and democratic development or the form of the electoral system.

35 BIBLIOGRAPHY

Adolfo, V. et al. (2012) Electoral violence in Africa. Policy notes (3). Uppsala: Nordiska Afrikainstitutet. [Online] Available from: http://www.diva- portal.org/smash/get/diva2:556709/fulltext01.pdf (Accessed 17 April 2017).

Amnesty International (2007) Burundi: No protection from rape in war and peace. [Online] Available from: https://www.es.amnesty.org/uploads/media/Burundi_SVAW_report_- _No_protection_from_rape_in_war_and_peace.pdf (Accessed 19 April 2017).

Ayiera, E. (2010) Sexual Violence in Conflict: a Problematic International Discourse. Feminist Africa. December (14). [Online] Available from: http://agi.ac.za/sites/agi.ac.za/files/feminist_africa_10.pdf (Accessed 27 April 2017).

Bardall, G. (2011) Breaking the Mold: Understanding Gender and Electoral Violence. [Online]. IFES White Paper. December. Available from: http://conflict.care2share.wikispaces.net/file/view/Gender_and_Electoral_Violence_2011.p df/297066496/Gender_and_Electoral_Violence_2011.pdf (Accessed 15 April 2017).

Bardall, G. (2015) Towards a More Complete Understanding of Election Violence: Introducing a Gender Lens to Electoral Conflict Research. In: 4th European Conference on Politics and Gender. [Online] Uppsala: ECGP Standing Group in Gender and Politics, pp. 1-25. Available from: https://ecpr.eu/Filestore/PaperProposal/ce51102c-24cb-4161-a3d8- ac125011be4a.pdf (Accessed 18 April 2017)

Barnett, M. (2005) Humanitarianism Transformed. Perspectives on Politics. 3 (4), pp. 723- 740.

BBC (2016) Burundi profile - Timeline. BBC News. 14 December. Available from: http://www.bbc.com/news/world-africa-13087604 (Accessed 6 April 2017).

Beach, D. & Pedersen, R. B. (2013) Process-tracing methods: foundations and guidelines. Ann Arbor: The University of Michigan Press.

Buchanan, E. (2015) Burundi: Who are the feared Imbonerakure youth? International Business Times. 3 June. [Online] Available from: http://www.ibtimes.co.uk/burundi-who- are-feared-imbonerakure-youth-1504301 (Accessed 27 January 2017).

36 Butler, C. K. et al. (2007) Security Forces and Sexual Violence: A Cross-National Analysis of a Principal Agent Argument. Journal of Peace Research. 44 (6), pp. 669–687.

Cockburn, C. (2004) ‘The continuum of violence: A gender perspective on war and peace’, in W. Giles & J. Hyndman (eds.) Sites of violence: Gender and conflict zones. Berkeley: University of California Press. pp. 24–44.

Cohen, D. K. et al. (2013) Wartime sexual violence: Misconceptions, Implications and Ways Forward. Washington DC: United States Institute of Peace. Special Report. [Online]. Available from: http://www.academia.edu/download/40184836/Cohen-Green- Wood_Wartime_Sexual_Violence_-_USIP_Feb_2013__1_.pdf

Daley, P. (2008) Gender & genocide in Burundi: the search for spaces of peace in the Great Lakes Region. African Issues. Oxford: James Currey.

Daley, P. (2014) ‘Unearthing the local: hegemony and peace discourses in Central Africa’, in F. McConnell et al. (eds.) The geographies of peace. London: IB Tauris.

Daley, P. & Popplewell, R. (2016) The appeal of third termism and militarism in Burundi. Review of African Political Economy. 43 (150), pp. 648–657.

Dijkman, N. et al. (2014) Sexual Violence in Burundi: Victims, perpetrators, and the role of conflict. HiCN Working Paper, March (172) [Online] Available from: http://www.hicn.org/wordpress/wp- content/uploads/2012/06/HiCN-WP-172.pdf (Accessed 9 May 2017).

Fisher, J. (2002) Electoral violence and conflict: A Strategy for Study and Prevention. [Online] IFES White Paper. February. Available from: http://unpan1.un.org/intradoc/images/docgifs/UNTC-small.gif (Accessed 8 May 2017).

Fjelde, H. & Höglund, K. (2016) Electoral Violence: The Emergence of a Research Field. APSA Comparative Democratization Newsletter, 14 (2), pp. 8-11.

Freedomhouse (2016) Freedom in the world 2016 - Burundi. [Online] Available from: https://freedomhouse.org/report/freedom-world/2016/burundi (Accessed 18 April 2017).

George, A. L. & Bennett, A. (2005) Case studies and theory development in the social sciences. BCSIA Studies in International Security. Cambridge/London: MIT Press.

Gerring, J. (2004) What Is a Case Study and What Is It Good for? The American Poltical Science Review. 98 (2), pp. 341–354.

37 Gottschall, J. (2004) Explaining wartime rape. Journal of Sex Research, 41 (2), pp. 129– 136.

Hafner-Burton, E. M. et al. (2014) When Do Governments Resort to Election Violence? British Journal of Political Science, 44 (1), pp. 149–179.

Hartill, L. (2017) A Vile Side of Ruling Party Youth League Members: Threats of Rape at Rally Caught on Video. [Online] Available from: https://www.hrw.org/news/2017/04/07/vile-side-ruling-party-youth-league-members (Accessed 13 April 2017).

Höglund, K. (2009) Electoral Violence in Conflict-Ridden Societies: Concepts, Causes, and Consequences. Terrorism and Political Violence, 21 (3), pp. 412–427.

Höglund, K. & Öberg, M. (2011) Understanding peace research: methods and challenges. London/New York: Routledge.

Hoofnagle, K. & Rothe, D. L. (2010) Overlooked and Overshadowed: The Case of Burundi. Critical Criminology, 18 (3), pp. 169–189.

HRW (2017) Burundi: Attacks by Ruling Party Youth League Members. [online]. Available from: https://www.hrw.org/news/2017/01/19/burundi-attacks-ruling-party-youth-league- members (Accessed 12 May 2017).

HRW (2016a) Burundi: Gang Rapes by Ruling Party Youth. [online]. Available from: https://www.hrw.org/news/2016/07/27/burundi-gang-rapes-ruling-party-youth (Accessed 13 January 2017).

HRW (2016b) Burundi: Intelligence Services Torture Suspected Opponents. [Online]. Available from: https://www.hrw.org/news/2016/07/07/burundi-intelligence-services- torture-suspected-opponents (Accessed 12 May 2017).

Huber, J. & Kammerud, L. (2016) Violence Against Women in Elections: An Excerpt from IFES’ Framework. IFES, August. [Online] Available from: http://www.ifes.org/publications/violence- against-women-elections-excerpt-ifes- framework (Accessed 28 April 2017).

Ikporr, I. (2017) Burundi youth call for rape of opposition. Cape Times. 13 April. [Online] Available from: http://www.pressreader.com/south-africa/cape- times/20170413/281509341051600 (Accessed 24 April 2017).

38 International Crisis Group (2016) Burundi: A Dangerous Third Term. Africa Report, May (235). Available from: https://d2071andvip0wj.cloudfront.net/235-burundi-a-dangerous- third-term.pdf (Accessed 3 February 2017)

King, G. et al. (1994) Designing social inquiry: scientific inference in qualitative research. Princeton New Jersey: Princeton University Press.

Krook, M. L. & Restrepo Sanin, J. (2016) Gender and political violence in Latin America: Concepts, debates and solutions. Política y gobierno, 23 (1), pp. 125–157.

Mbabazi, D. (2015) Winning the war against Gender Based Violence. The New Times, 26 November. [Online] Available from: http://www.newtimes.co.rw/section/article/2015-11- 26/194732/ (Accessed 5 May 2017).

McNeish, H. (2016) Running from Rape in Burundi. [Online]. Available from: http://tracks.unhcr.org/2016/05/running-from-rape-in-burundi/ (Accessed 24 April 2017).

Meger, S. (2010) Rape of the Congo: Understanding sexual violence in the conflict in the Democratic Republic of Congo. Journal of Contemporary African Studies, 28 (2), pp. 119– 135.

Murunga, G. R. (2011) Spontaneous or premeditated: post-election violence in Kenya. Discussion Paper (57). Uppsala: Nordiska Afrikainstitutet.

Muvumba Sellström, A. (2015) Stronger than Justice: Armed Group Impunity for Sexual Violence. PhD. Report, Department of Peace and Conflict Research; 104. Uppsala: Department of Peace and Conflict Research.

Nordås, R. (2011) Sexual Violence in African Conflicts. CSCW Policy Brief (1). [Online] Available from: https://www.usip.org/sites/default/files/missing-peace/Ragnhild- Nordas.pdf

OHCHR (2017) Grotesque rape chants lay bare campaign of terror by Burundi militia – Zeid. [Online] Available from: http://ohchr.org/EN/NewsEvents/Pages/DisplayNews.aspx?NewsID=21505&LangID=E.

Pézard, S. & De Tessières, S. (2009) ‘Insecurity is also a war’ an assessment of armed violence in Burundi. Geneva: Geneva Declaration Secretariat. [Online] Available from: http://www.genevadeclaration.org/fileadmin/docs/Geneva-Declaration-Armed-Violence- Burundi-EN.pdf (Accessed 15 April 2017).

39 Pruitt, L. (2012) Looking Back, Moving Forward: International Approaches to Addressing Conflict-Related Sexual Violence. Journal of Women, Politics & Policy, 33 (4), pp. 299– 321.

Sanghani, R. (2016) ‘I became a child refugee to escape rape. It didn’t work’. The Telegraph. 12 October. [Online] Available from: http://www.telegraph.co.uk/women/life/i-thought-i- was-safe---i-was-wrong-the-cruel-twist-of-fate-that/ (Accessed 24 April 2017).

Steers, J. (2016) Women Say They Are Being Raped as Part of the President of Burundi’s Fight to Keep Power. Time. 14 January. [Online] Available from: http://time.com/4179101/rape-burundi/ (Accessed 24 April 2017).

Thomas, K. et al. (2013) Political transition and sexual and gender-based violence in South Africa, Kenya, and Zimbabwe: a comparative analysis. Gender & Development, 21 (3), pp. 519–532.

UNFPA (1999) Reproductive Health in Refugee Situations: An Inter Agency Field Manual (UNHCR/WHO/UNFPA) [Online] Available from: http://www.unhcr.org/publications/operations/3bc6ed6fa/reproductive-health-refugee- situations-inter-agency-field-manual-unhcrwhounfpa.html (Accessed 8 May 2017).

UNIIB (2016) Report of the United Nations Independent Investigation on Burundi (UNIIB) established pursuant to Human Rights Council resolution S-24/1. [Online] Available from: http://reliefweb.int/sites/reliefweb.int/files/resources/A_HRC_33_37_E_AUV_.pdf (Accessed 12 April 2017).

United Nations News Service (2017) Grotesque rape chants expose ‘campaign of terror’ in Burundi – UN rights chief [Online] Available from: http://www.un.org/apps/news/story.asp?NewsID=56574#.WQzL7FKHK8U (Accessed 5 May 2017).

Uvin, P. (2010) Structural causes, development co-operation and conflict prevention in Burundi and Rwanda. Conflict, Security & Development, 10 (1), pp. 161–179.

Wanjiku Kihato, C. (2015) ‘Go Back and Tell Them Who the Real Men Are!’: Gendering Our Understanding of Kibera’s Post-election Violence. International Journal of Conflict and Violence. 9 (1), pp. 13-24

Wood, E. J. (2006) Variation in Sexual Violence during War. Politics & Society, 34 (3), pp. 307–342.

40 Wood, E. J. (2009) Armed Groups and Sexual Violence: When Is Wartime Rape Rare? Politics & Society, 37 (1), pp. 131–161.

Wood, E. J. (2014) Conflict-related sexual violence and the policy implications of recent research. International Review of the Red Cross, 96 (894), pp. 457–478.

Yin, R. K. (1981) The Case Study Crisis: Some Answers. Administrative Science Quarterly, 26 (1), pp. 58-65.

41 APPENDIX A.

Causes of electoral violence

Conditions enabling the use of election Factors triggering electoral violence violence

Nature of - Patrimonial politics - Violent actors politics - Conflict cleavages participating in elections - Violence as a legitimate political tool - Biased police - Culture of impunity - Access to arms

Nature of - Competitiveness - Misuse of political rights elections - Political mobilisation - Militant mobilisation - Stakes - “Close races” - Exposure to attacks

Electoral - Systems creating clear winners and - Political usage of institutions losers electoral administration - Few regulations about electoral - Election fraud conduct - Unwanted or unexpected - Administration with few checks and outcome of the election little power

(Höglund, 2009, p.423)

42 APPENDIX B.

Key events surrounding the 2015 electoral crisis

Date Event Agathon Rwasa, leader of former rebel group FNL, announces decision to run in 2015 August 2013 presidential election after three years in exile. Parliament blocks an attempt from CNDD-FDD government to change the constitutions March 2014 power-sharing agreement between ethnic groups. UN senior official is ordered to leave Burundi after UN report is released, warning that the April 2014 government is arming Imbonerakure supporters in preparation for the upcoming elections. Nkurunziza announces his decision to stand for a third term. April 2015 Protestors take the streets in Bujumbura. The Constitutional Court rules in favour of Nkurunziza’s third term despite reports of intimidation of judges. May 2015 A group of military leaders lead a failed coup attempt when Nkurunziza is in Tanzania. Oppositional UPRONA leader Zedi Feruzi is shot dead. Presidential elections are postponed until July and parliamentary elections until the end of June after a month of protests. June 2015 Opposition leaders and some leaders in the regime flee Burundi after voicing opposition to the third term. Nkurunziza wins the election with 70 % of the vote. CNDD-FDD wins parliamentary July 2015 election. UN observers announce that elections were neither free, nor credible. African Union announces plans to send 5000 troops to protect civilians from escalating January 2016 violence. Nkurunziza threatens to counter the plans. March 2016 The EU suspends its financial aid to the Burundian government. A UN report accuses Rwanda of supporting Burundian rebels. May 2016 Kagame denies the allegations. HRW releases report about gang-rape by the Imbonerakure and escalation of sexual July 2016 violence. Burundi rejects the deployment of UN police to end political violence, refereeing to a August 2016 violation of Burundian sovereignty. Burundi formally signs a bill into law about withdrawal from the International Criminal October 2016 Court. UN High Commissioner fro Human Rights, Zeid Ra’ad Al Hussien expresses deep concern April 2017 for developments of sexual violence in Burundi.

Sources: (BBC, 2016; Freedomhouse, 2016; HRW, 2016a; UNIIB, 2016; OHCHR, 2017)

43